Prospective Research Trend Analysis on Zero-Energy Building (ZEB): An Artiﬁcial Intelligence Approach

: While global attention to zero-energy building (ZEB) has surged as a sustainable coun-termeasure to high-energy consumption, a congruent expansion in research remains conspicuously absent. Addressing this lacuna, our study harnesses public research and development grant data to decipher evolving trajectories within ZEB research. Distinctively departing from conventional methodologies, we employ state-of-the-art natural language processing (NLP) artiﬁcial intelligence models to meticulously analyze grant textual content pertinent to ZEB. Our ﬁndings illuminate an expansive spectrum of ZEB-related research, with a pronounced focus on the holistic continuum of energy supply, demand, distribution, and actualization within architectural conﬁnes. Theoretically, this work delineates key avenues ripe for future empirical exploration, fostering a robust academic foundation for subsequent ZEB inquiries. Practically, the insights derived bear signiﬁcant implications for practitioners, informing optimal implementation strategies, and offering policymakers coherent roadmaps for sustainable urban development. Collectively, this study affords a panoramic perspective on contemporary ZEB research contours, enhancing both scholarly comprehension and practical enactment in this pivotal domain.


Introduction
As global efforts to address climate change and pursue sustainable development intensify, zero-energy building (ZEB) is increasingly considered a viable solution for reducing carbon dioxide (CO 2 ) emissions and minimizing energy consumption within the building, construction, and architecture sector [1,2].In many developed countries, the building sector accounts for 30% to 40% of total energy consumption and a sizable proportion of greenhouse gas emissions [3][4][5][6][7][8][9][10]. Consequently, ZEB has become a promising alternative many countries have actively pursued through technology and policy initiatives to improve building energy efficiency [1,11].Similarly, the building sector accounts for over a quarter of Korea's total energy consumption.Recognizing the need to address climate change, reduce energy demand, and decrease greenhouse gas emissions, the Korean government has prioritized promoting ZEB to foster new industries and technological advancements [12,13].
In conceptual terms, ZEB refers to a building that significantly enhances energy efficiency by incorporating renewable energy sources and minimizing energy losses [6].In practice, however, ZEB is a complex concept with definitions that vary depending on the emphasis on specific aspects of building management technology.Moreover, the notion of a zero-energy state and how to measure the energy balance of a building remain ambiguous, with no clear consensus [1,11,14,15].Consequently, a comprehensive understanding of the technical elements and research related to ZEB is still needed [14].
However, this does not imply a lack of research aimed at clarifying the definition of ZEB and calculating and evaluating various related parameters.For instance, Marszal et al. [1] reviewed multiple research papers on ZEB, presented definitions of ZEB and energy calculation techniques in lieu of conventional scientometric analysis stems from several considerations.Firstly, a plethora of studies have already employed scientometric or bibliometric methodologies to probe into ZEB-related research and development trends [29][30][31].Adopting a similar approach would, thus, dilute the novelty of this investigation.Secondly, recent advancements in NLP, particularly methodologies harnessing architectures like BERT and Transformer, have demonstrated remarkable efficacy.Consequently, this study posits that leveraging these cutting-edge AI methodologies for analyzing ZEB-related R&D trends can yield more granular and empirically robust insights.
The research questions accordingly are as follows: (1) How does the trend analysis of ZEB research using R&D grant data differ from those derived from scientific publication data?(2) What technological opportunities can be observed for the future-oriented directions of ZEB-related R&D using grant data from major countries?(3) What is the current status of global ZEB R&D based on the analysis of R&D grant data?(4) How does the knowledge structure of ZEB research manifest when evaluated through R&D grant data?(5) Which future research directions emerge when R&D grant data for ZEB is analyzed through NLP based on AI models?
From these research questions, the research objectives presented in this paper are as follows: (1) To conduct a comprehensive trend analysis of ZEB research using R&D grant data from major countries.(2) To contrast the insights derived from R&D grant data with those typically obtained from scientific publication data.(3) To pinpoint technological opportunities that elucidate future-oriented directions in ZEB-related R&D.(4) To encapsulate the present status of global ZEB R&D by examining invested grant data.(5) To identify and map out the knowledge structure within the ZEB research domain.(6) To leverage an innovative methodology employing NLP based on AI models for analyzing R&D grant document data, moving away from conventional bibliometric or scientometric methods.

Data Collection and Preprocessing
This section outlines the steps for obtaining R&D grant datasets related to ZEB.The task involved creating a query set that includes the concepts of ZEB and NZEB.The R&D grant data related to ZEB were restricted to projects from 2000 to 2022.Data collection post-2000 was strategically chosen to discern shifts in R&D trends encompassing early ZEB-related research up to contemporary advancements.This decision was made in light of the fact that the Kyoto Protocol, an important international treaty aimed at mitigating global warming, was adopted in 1997.
The database used to collect ZEB-related R&D grant data is Dimensions.ai, an integrated research information system provided by Digital Science (https://www.dimensions.ai, accessed on 24 July 2023).This system was chosen due to its ability to organize and provide considerable global R&D grant data systematically.The research category feature offered by Dimensions.ai was used to narrow down data collection to research fields relevant to ZEB to reduce noise in the collected data.During the data collection process, the Australian and New Zealand Standard Research Classification 2020 (ANZSRC 2020) was used to limit research areas based on research field codes.The ANZSRC 2020 is a commonly used statistical classification system for measuring and analyzing R&D activities in Australia and New Zealand.The query set and parameters presented below were used to extract only ZEB-related data from the R&D grant data provided by Dimensions.ai.The asterisk indicates a fuzzy search.A total of 3456 documents were retrieved.

•
The data date range was 2000 to 2022.

•
Only documents of the grant type were used.

•
Duplicated data were removed based on the Grant ID.

•
The query set is as follows:  the data collection process, the Australian and New Zealand Standard Research Classification 2020 (ANZSRC 2020) was used to limit research areas based on research field codes.The ANZSRC 2020 is a commonly used statistical classification system for measuring and analyzing R&D activities in Australia and New Zealand.The query set and parameters presented below were used to extract only ZEB-related data from the R&D grant data provided by Dimensions.ai.The asterisk indicates a fuzzy search.A total of 3456 documents were retrieved.

•
The data date range was 2000 to 2022.

•
Only documents of the grant type were used.

•
Duplicated data were removed based on the Grant ID.

•
The query set is as follows:

Data Analysis
Many studies previously conducted employed scientometrics to identify the domain of scientific knowledge using quantitative analysis methodologies, such as coauthorship, cocitation, keyword co-occurrence, and cluster analysis, facilitating the exploration of hidden implications and the identification of innovative research areas [35][36][37][38].However,

Data Analysis
Many studies previously conducted employed scientometrics to identify the domain of scientific knowledge using quantitative analysis methodologies, such as coauthorship, cocitation, keyword co-occurrence, and cluster analysis, facilitating the exploration of hidden implications and the identification of innovative research areas [35][36][37][38].However, to identify more practical research fields and content related to ZEB, this study employs an AI-based clustering analysis after conducting NLP on the unstructured ZEB-related R&D grant data, specifically the titles and abstracts.

Document Embedding
The initial step in analyzing R&D grants related to ZEB is converting each document into numerical data, called embedding in NLP.In recent years, pretrained language models have become commonplace in the embedding process, and this study employs the widely used bidirectional encoder representations from transformers (BERT) model [39].As BERT produces distinct embeddings based on word context, it is a suitable embedding method for comprehending ZEB-related research content.Moreover, many pretrained models are available as open sources, making them readily accessible for analysis.Numerous techniques can generate BERT embeddings with text data.This study employs Python and the sentence-transformers package to generate BERT embeddings for ZEB-related R&D grant documents.The sentence-transformers package is acknowledged for producing high-quality document-level embeddings [40,41], making it a suitable tool for this research.This study converts the documents into 512-dimensional numerical data via the BERT embedding.

Dimension Reduction and Document Clustering
A clustering analysis process is necessary to group documents sharing similar topics into clusters.However, numerous clustering algorithms struggle to manage high dimensions effectively, making it imperative to reduce the dimension of embedding beforehand.
Of the various dimensionality reduction algorithms available, uniform manifold approximation and projection (UMAP) [42] is recognized for its efficacy in preserving a significant proportion of high-dimensional local structures in low dimensionality.This study employs the UMAP algorithm for dimensionality reduction by installing the umaplearn package from Python.Through the UMAP algorithm, we reduced the dimension size to five while maintaining the local neighborhood size at 15.If the dimension is excessively low, pertinent information may be lost, whereas if the dimension is overly high, the clustering result may be suboptimal.Thus, we adopted the parameters suggested by the umap-learn package.
After reducing the document embedding dimension to five, we used the hierarchical density-based spatial clustering of applications with the noise (HDBSCAN) algorithm to cluster documents [43,44].The HDBSCAN is a density-based clustering algorithm that synergizes well with UMAP because UMAP significantly preserves local structures, even in a low-dimensional space [45].Moreover, HDBSCAN is advantageous because it does not compel data points to belong to clusters, as it considers some data points to be outliers [46].We installed the hdbscan package of Python to employ this algorithm, as with the previous algorithms.
This process allows similar documents to be grouped to form clusters.Furthermore, reducing the dimension size to two enables the visualization of the cluster analysis result on a two-dimensional plane, while unclustered outliers can be visualized separately.In cases where the number of analyzed clusters is large, the clusters may not be accurately represented on a plane.Nonetheless, reducing the dimensionality to two can still reveal local structures in most cases.

Topic Modeling
This study applied the BERTopic algorithm to identify topics within clusters.The BERTopic technique extracts topics from text embeddings using a language model, such as BERT [47].The process was implemented by installing the BERTopic package in Python, which is modular and uses a series of steps to create a topic model.
Steps 1 to 3 are identical to the document clustering process described earlier, except that, in topic modeling, the process is performed within the clusters generated in the previous step.In Step 4, to generate topics without assuming any expected structure of the clusters, BERTopic employs a bag-of-words approach by counting the frequency of each word in each cluster.After generating the word frequency representations in Step 4, Step 5 uses class-based term frequency-inverse document frequency (c-TF-IDF) to determine how one cluster differs from another.For instance, it calculates the importance of words within clusters and identifies which words are common in Cluster 1 but not in other clusters.The following equation can be used to calculate c-TF-IDF: The process of collecting and analyzing ZEB-related R&D grant data in this study is illustrated in Figure 2.
each word in each cluster.After generating the word frequency representations in Step 4, Step 5 uses class-based term frequency-inverse document frequency (c-TF-IDF) to determine how one cluster differs from another.For instance, it calculates the importance of words within clusters and identifies which words are common in Cluster 1 but not in other clusters.The following equation can be used to calculate c-TF-IDF: The process of collecting and analyzing ZEB-related R&D grant data in this study is illustrated in Figure 2.

Descriptive Analysis
This section presents the results of a descriptive statistical analysis of the R&D grant data related to ZEB collected from Dimensions.ai.Between 2000 and 2022, the United Kingdom (UK) funded the most ZEB-related R&D grants, with 719 grants being funded, followed by the United States (US), Canada, Belgium, and China funding 714, 474, 384, and 265 grants, respectively.Most of the grants funded in Belgium account for numerous projects funded by the European Commission (EC) because Belgium is the seat of the EC (Table 1).

Funder Country
The Number of Funded R&D Grants United Kingdom 719 United States 714

Descriptive Analysis
This section presents the results of a descriptive statistical analysis of the R&D grant data related to ZEB collected from Dimensions.ai.Between 2000 and 2022, the United Kingdom (UK) funded the most ZEB-related R&D grants, with 719 grants being funded, followed by the United States (US), Canada, Belgium, and China funding 714, 474, 384, and 265 grants, respectively.Most of the grants funded in Belgium account for numerous projects funded by the European Commission (EC) because Belgium is the seat of the EC (Table 1).Table 3 summarizes the average funding for ZEB-related R&D grants during the study period.The investment amounts were converted into US dollars from each country's respective currency.However, the average funding in Belgium includes grants funded by the EC; thus, it is not a good representation of the funding in Belgium alone.Japan is the top country for funding, followed by Belgium, New Zealand, the UK, and Czechia regarding the average investment for ZEB-related R&D grants.Although New Zealand and Czechia did not rank high in the number of R&D grants, the funding per grant is high given their high average investments.

Document Clustering Results
Following the data analysis method, we embedded the text using BERT with the title and abstract of the R&D grants related to ZEB, reduced the dimensionality using the UMAP algorithm, and performed clustering using the HDBSCAN algorithm.Consequently, the documents were grouped into 25 clusters.
Figure 3 presents the results of the document embedding and clustering, reduced to two dimensions.The shaded areas correspond to outliers that did not form clusters.After dimensionality reduction from 512 to five dimensions using the UMAP algorithm, it was possible to identify the clustered structure of similar documents, even when the dimension size was reduced to two for visualization.The size of each cluster is listed in Table 4.
Following the data analysis method, we embedded the text using BERT with the title and abstract of the R&D grants related to ZEB, reduced the dimensionality using the UMAP algorithm, and performed clustering using the HDBSCAN algorithm.Consequently, the documents were grouped into 25 clusters.
Figure 3 presents the results of the document embedding and clustering, reduced to two dimensions.The shaded areas correspond to outliers that did not form clusters.After dimensionality reduction from 512 to five dimensions using the UMAP algorithm, it was possible to identify the clustered structure of similar documents, even when the dimension size was reduced to two for visualization.The size of each cluster is listed in Table 4.

Results of Topic Modeling and Content Analysis by Clusters
As mentioned, topic modeling was performed for each cluster using the BERTopic algorithm.The number of topic groups extracted per cluster varied depending on cluster size.Figure 4 presents the results of topic modeling for the 25 clusters.this study, we organized clusters derived from a comprehensive analysis of R&D grants.Utilizing the BERTopic model, a state-of-the-art topic modeling technique, we extracted key thematic elements to inform the titles of each cluster.These titles were formulated based on keyword prominence and their relevance to the overarching themes of the respective R&D grants.The topics by cluster are the representative research content on ZEB-related R&D gathered in each cluster.This study examined the actual R&D grant content for each cluster by identifying the research content centered on the cluster topic.The results of identifying the research content for each cluster are presented in Table 5.For example, the present thesis posits that the R&D grants attributed to Cluster 0 are primarily dedicated to conducting R&D endeavors on advanced nuclear technology.Moreover, these grants emphasize advancing the development of sustainable ZEB by incorporating innovative nuclear reactor designs, advanced fuel assemblies, and state-ofthe-art diagnostic and monitoring tools.
The R&D grants belonging to Cluster 1 focus on developing advanced materials, innovative structural systems, and seismic design methods for enhancing the seismic safety and energy efficiency of buildings.These research studies aim to optimize the thermal and acoustic performances of masonry housing and develop sustainable and low-risk structural buildings using high-strength materials, timber composites, and energy-dissipating elements.These studies also investigate the response characteristics of super-high-rise structures under long-term ground motion and establish damping control mechanisms to improve their seismic performance.
The field of studies in Cluster 2 focuses on the dynamics and transport of fluids in complex systems, including developing advanced numerical methods and models for simulating turbulent flows and investigating the influence of numerous factors, such as rough surfaces, pressure gradients, and interfacial area concentration on the behavior of fluids.This work encompasses research areas, such as aerodynamics, hydrodynamics, and atmospheric physics, and applications in energy, urban air quality, and material science.Table 6 similarly summarizes the research areas and content of the 25 clusters.In this study, we organized clusters derived from a comprehensive analysis of R&D grants.Utilizing the BERTopic model, a state-of-the-art topic modeling technique, we extracted key thematic elements to inform the titles of each cluster.These titles were formulated based on keyword prominence and their relevance to the overarching themes of the respective R&D grants.These R&D grants cover a wide range of research related to seismic safety, energy efficiency, and sustainability of building structures.The grants explore different approaches to improve the seismic resistance of building structures, such as the use of novel materials, advanced design methods, and innovative structural systems.Some grants focus on upgrading energy and the sustainability of masonry buildings, whereas others investigate the seismic behavior and performance of high-rise structures.Additionally, grants are dedicated to developing new seismic isolation and vibration control technologies, such as self-centering buckling restrained braces and high-damping elastomers.These grants aim to improve the environmental performance of building materials, such as the development of sustainable ceramic brick masonry veneer walls for building envelopes.
Cluster 2 Fluid dynamics and turbulence modeling for ZEB These R&D grants encompass research fields related to fluid dynamics and mathematical physics, focusing on turbulence, flow dynamics, and modeling.Specific research topics include the study of rough-walled turbulent flows, developing anti-icing materials, investigating interfacial area concentration transport in bubbly flows, effects of surfactants on drag reduction, and constructing vortex-wave-based turbulence models.

Cluster 3
Advanced photovoltaic technologies and integration for ZEB These R&D grants cover various research fields on developing more efficient and sustainable photovoltaic technologies.These research areas include the development of high-transparency, high-conductivity spectrally selective coatings, solution-processed inorganic thin-film photovoltaic devices, atomically thin photovoltaics, organic semiconducting materials, concentrated solar energy storage, and solar energy storage into redox flow batteries.The grants also encompass developing sustainable materials and manufacturing processes, designing efficient solar cells with a low CO 2 footprint, and integrating solar panels into building materials.Other research areas include developing defect-tolerant photovoltaic materials, interfacial engineering of photovoltaic devices, and integrating radiative cooling into photovoltaic/thermal panels in buildings.These R&D grants relate to research fields aimed at reducing carbon emissions of cementitious products and developing low-carbon alternatives to traditional concrete for a net-zero future.Research includes developing novel additives and technology to enhance the performance of low-carbon cement, valorizing waste materials (e.g., contaminated waste glass and steel slag) for use in construction, and producing zero-emission and low-cost concrete materials using bio-catalytic calcium carbonate cementation and ultralow binder content.Other projects focus on the life-cycle assessment of sustainable cement, using IoT, machine learning, and big data to transform the cement supply chain, and testing and reusing concrete-encased steel from the 1950s.

Cluster 5 Sustainable production and advanced tech integration
These R&D grants cover various topics related to sustainable production, energy efficiency, and advanced technology development.The grants support research in sustainable cement production, low-emission transport, ultralow-cost sensors, energy storage, embedded systems, and communication systems infrastructure.

Cluster 6 Geothermal energy and thermal storage systems
These research grants cover various topics related to geothermal energy, thermal energy storage, and mathematical modeling for the energy-efficient design of underground systems.The grants aim to advance the development of sustainable energy systems for ZEB, including designing and optimizing geothermal heat pump systems, thermal energy storage, and underground mine ventilation systems.

Cluster 7
Power electronics in ZEB These R&D grants focus on various aspects of power electronics and their applications in ZEB.The research topics include developing high-performance power converters using new materials (e.g., silicon carbide (SiC)), analyzing and modeling switching arcs, designing switching-cell-array-based power electronics for electric vehicles, developing smart switching devices for energy-saving applications, and investigating efficient and power-dense modular power electronic architectures for utility-scale DC-AC conversion.

Cluster 8 Energy generation, storage, and conversion technologies
These R&D grants cover diverse topics related to energy generation, storage, and conversion.Some of the research areas include wind turbine technology, laminar flow seals, green air transportation, ultralow wind-speed wind power generation, active sensor technologies, energy storage technology for power grids and micro-grids, electromechanical energy conversion, propeller aerodynamic interaction and noise characteristics, resonant gyro micro hemispherical concave arrays, nutating disk engines, linear synchronous permanent magnet motors, high-speed generators, mathematical modeling of nonlinear effects in electrical systems, spacecraft flywheel energy storage, energy-efficient control algorithms for advanced aircraft, robust control of stochastic delay Hamiltonian systems, DC-saturation-relieving contra-rotating wind energy conversion systems, and light pressure for space propulsion.These grants aim to improve the efficiency, reliability, and sustainability of energy systems for various applications.

Cluster 9
Ultralow power electronic circuits and hardware These R&D grants focus on energy-efficient hardware and circuits for ultralow power consumption, including efficient processor design, memory and digital circuit boundary exploration, ultralow voltage SRAM architectures, adiabatic circuits, cryogenic adiabatic CMOS, and of analog integrated system development for ultralow voltage applications.The grants also cover developing novel capacitor-less dynamic random access memory technology, very low-power system-on-a-chip super dynamic voltage regulator key technology research, and high-bandwidth sensing for wide-bandgap power conversion.The listed R&D grants cover research on developing energy-efficient and ultralow power devices, systems, and technology for different applications, including building control, wireless communication, sensing systems, and Internet of Things (IoT) devices.The research content encompasses designing and optimizing digital and analog circuits, memory architectures, micro-electromechanical devices, energy harvesting systems, wireless sensors, antennas, and communication protocols for ultralow power and zero-power operation.
Cluster 11 Nanotechnology and advanced material systems for ZEB The listed R&D grants explore the use of molecular conformational dynamics for electromechanical qubits, semiconducting carbon nanotube polaritonic devices, and graphene spintronics.Additionally, the grants cover developing solid-state electrolytes for all-solid thin-film Li-ion batteries, 3D nanophotonic devices, and particulate-based functional macromolecules, among others.

Cluster 12 Advanced materials and technologies for ZEB and transportation
The listed R&D grants cover research on ZEB.Some of the research areas include developing disruptive polyurethane foams with improved passive fire protection, pressure-efficient hydrogen storage, preform technology for automotive part production, material science, digital materials, ultralow wear coatings, ultra-lightweight clay-aerogel materials, electromechanical formation, hybrid lightweight foam cores, gradient structures for flexible components, vacuum insulation panels, multiscale investigation and mimicry of naturally occurring composite materials, and fire protection systems for munitions.These research projects aim to improve energy efficiency, safety, insulation, and lightweight building materials and vehicles and develop advanced materials and coatings.

AI-driven marine ecosystem interactions and sustainable energy monitoring
These grants focus on researching and developing monitoring systems, applying AI and computer models, optimizing sustainable energy production, restoring ecosystems, and building capacity for sustainable interactions with marine ecosystems.

Cluster 14 Zero-emission solutions for sustainable transportation
The listed R&D grants cover various fields related to ZEB and sustainable transportation.The research content encompasses developing and optimizing smart diesel fuel solutions; zero-carbon power solutions for ships; renewable fuel range extenders; clean fuel supply solutions; climate effects reduction of construction materials; electrified road transports; and zero-emission vehicles, marine vessels, and machinery.These projects also involve feasibility studies, data analytics tools, and life-cycle modeling to support the transition to a sustainable zero-emission transportation system.

Cluster 15
Hydrogen solutions and carbon management for ZEB These R&D grants are related to various ZEB research fields, including clean hydrogen production, hydrogen storage and transportation, carbon capture and utilization, energy efficiency, and fuel cell technology.The research content encompasses a range of topics, such as developing efficient and flexible hydrogen production methods, assessing hydrogen embrittlement in pipelines, exploring novel hydrogen-resistant materials, and improving fuel cell technology for low-emission transportation and remote site energy production.These R&D grants are related to renewable energy generation from water, such as tidal, wave, and hydroelectric power and technology for energy storage and distribution.The research topics cover flow control, system design and optimization, reliability and risk management, control strategies, and numerical modeling and simulation.The grants also focus on developing novel devices and prototypes, such as the PAX rotor and the SOURCE hydropanel, for sustainable energy generation and supply.

Cluster 18
Sustainable ZEB solutions in water treatment and industrial manufacturing These R&D grants focus on achieving ZEB.The covered topics include carbon capture and utilization in biomanufacturing, sustainable waste management solutions, energy-neutral wastewater treatment, decentralized water technology, zero-carbon concrete production, low-energy water treatment processes, and efficient nitrogen removal from wastewater.Other research areas include developing innovative membranes, using bioreactors and biofilms for water treatment, and exploring new manufacturing processes and materials for building envelopes.Some grants address the challenges facing specific industries, such as glass manufacturing and textile production, whereas others focus on water disinfection and contaminated groundwater remediation.

Cluster 19
Advanced window systems for energy-efficient buildings These research grants focus on developing technology and materials for improving the energy efficiency of buildings through advanced window systems, including ultra-thin glass membranes, polyurethane window systems, smart windows with high thermal and acoustic insulation, affordable high-performance windows, and lightweight switchable smart solutions.Other projects aim to develop new framing technology for highly insulating glazing, adaptable envelopes for building refurbishment, and water flow glazing systems.The goal is to reduce building energy consumption and promote sustainable materials and manufacturing processes.

Cluster 20 Retrofitting and energy optimization in existing buildings
These R&D grants are related to retrofitting existing buildings to improve energy efficiency and move toward ZEB.The research fields include technology development, manufacturing processes, smart textiles, energy management systems, modular and versatile process units, performance analysis, machine learning, life-cycle assessment, indoor climate and energy performance, and profitability analysis.
Cluster 21 Wood-centered approaches for ZEB and sustainable constructions These R&D grants focus on ZEB using wood as a primary material.The research can include designing ultralow energy green buildings with renewable wood materials, enhancing the collapse resistance of cross-laminated timber buildings, developing finishing and densification solutions for interior wood products, creating affordable zero-carbon constructions, developing sustainable solutions for structural floor systems, optimizing transformation processes and machine tools for wood, creating new wood-fiber panels for wood buildings, achieving net-zero energy and carbon in wooden buildings, developing bulk insulation from cedar transformation co-products, and stabilizing wood in a circular economy.

ZEB technologies and building-integrated renewable systems
The listed R&D grants cover diverse ZEB fields, including envelope material systems, building retrofits, renewable energy technology, energy optimization in communities, smart energy management systems, circular economies, decarbonization of energy systems, and modeling and optimization of building integrated renewable energy systems.The grants also cover various types of technology, such as air source heat pumps, solar thermal and photovoltaic systems, storage systems, gas networks, and regenerative high-performance curtain walls.
Cluster 24 Advanced energy systems and sustainable building innovations These R&D grants relate to various aspects of zero energy building, including design and control of advanced energy systems, digitalization of power and energy systems, hybrid air conditioning, grid energy storage, sustainable energy systems, anaerobic digestion, thermochemical energy storage, ground-source heating and cooling, and heat recovery ventilation.Other topics include innovative energy-saving devices, low-carbon heating and cooling systems, and microbial contamination in energy-saving ventilation equipment.The grants also cover optimizing thermal energy storage, effectively using renewables, and employing energy-efficient building materials, such as radiative cooling paints.

Discussion
The clustering results highlight the diverse range of ZEB research, covering a spectrum from advanced nuclear technology to fluid dynamics in complex systems.Document categorization into 25 distinct groups signifies the breadth of subjects under the purview of ZEB research.Cluster 0 points to a nascent inclination towards harnessing advanced nuclear technology for sustainable ZEB outcomes.The focus on innovative reactor designs and advanced diagnostic tools alludes to a potential pivot towards nuclear energy as a sustainable solution for zero-energy buildings.This warrants an in-depth exploration of its feasibility, associated ramifications, and public perception.Cluster 1 accentuates the relevance of innovations in material science and the imperative of seismic safety in ZEB.Given the mounting concerns over environmental calamities, there is an increased emphasis on seismic design techniques and ensuring the resilience of towering structures.It is essential to evaluate how such advancements might redefine the established architectural and engineering paradigms in ZEB.The attention to fluid dynamics, as highlighted in Cluster 2, is of particular interest.When explored in the context of aerodynamics and atmospheric physics, fluid dynamics can profoundly impact building design, ventilation strategies, and energy conservation approaches.Delving into the interplay between these elements and ZEB design promises fresh perspectives.
The employment of AI tools, like BERT and UMAP, facilitates a nuanced exploration of ZEB-focused R&D trends.Yet, it remains critical to reflect on potential biases, ascertain the results' reliability, and acknowledge the limitations of this AI-driven approach.The ungrouped outliers could represent untapped knowledge, potentially pointing to emerging or niche research areas on the verge of broader recognition.An in-depth analysis of these outliers could delineate emerging directions in ZEB research.Using AI to investigate prospective research trends in ZEB has demarcated clear clusters echoing the field's multifaceted nature.Each cluster epitomizes a distinct aspect of ZEB, shedding light on the intricate avenues of sustainable building design research.Nevertheless, while these clusters serve as a valuable guide to current ZEB research, inherent limitations and potential areas of deeper inquiry emerge.Despite BERT's proven effectiveness in text embedding, biases innate to its pre-trained model can creep in.Moreover, while UMAP is proficient at dimensionality reduction, its sensitivity to hyperparameters might skew the clustering outcome.Furthermore, HDB-SCAN's capability in capturing diverse density clusters might occasionally miss more diffuse ones, categorizing certain research areas as outliers.
The study's focus on only the titles and abstracts of R&D grants may inadvertently neglect subtle nuances or emergent themes present in the full text.Additionally, as this study provides a snapshot of ZEB trends, it may not trace the entire thematic evolution, especially emerging or waning research facets.Outliers, which do not neatly fit into the predefined 25 clusters, could be hinting at avant-garde, cross-disciplinary, or specialized research trajectories.An exhaustive qualitative probe into these outliers might unearth pioneering ZEB research directions.
Undertaking a time-based examination of the R&D grants might illuminate the temporal evolution of ZEB research.Such an inquiry can chronicle the birth, growth, and possible waning of distinct research themes, furnishing a fluid overview.Subsequent studies stand to gain from extending their scope to the full text of R&D grants, thereby ensuring a more holistic grasp of the research nuances.Complementing this with external datasets, like citation networks or patent databases, would bestow a comprehensive perspective on the ZEB research's impact and innovative pathways.

Conclusions
Energy consumption in the building and construction sectors remains a salient challenge across many nations.Amid the global agreement to transition from fossil fuels to renewable energies, Zero-Energy Building (ZEB) stands out as a viable alternative, garnering extensive research attention [1,2].This research introduced an AI-driven methodology to examine ZEB-centric R&D grants, aiming to decipher future trajectories and enrich our understanding of the discipline.
Diverging from conventional scientific analyses that primarily focus on academic articles, our study prioritized R&D projects to illuminate the prospective avenues of ZEB research.Despite inherent challenges in analyzing all ZEB-related R&D undertakings due to data accessibility constraints, such a methodological choice is pivotal.These projects often epitomize national R&D agendas, overseen by governmental or public entities.Our analysis emphasized concerted efforts to amplify ZEB efficiency, elevate photovoltaic performance, and blueprint smart cities integrating ZEB with transportation and avant-garde technologies, such as ICT and sensors.These insights resonate with Rotolo et al. [36], underscoring the relevance of R&D funding data in spotlighting imminent research directions.The value proposition of our study lies in its novel methodology to envisage the ZEB research horizon.By leveraging R&D project data, we offer a holistic vantage point and insights distinct from those gleaned through traditional article analyses.Moreover, this endeavor affirms the potency of AI techniques as instrumental scientific apparatuses, thereby extending the frontiers of such inquiries.
Our findings can serve both theoretical and applied facets, assisting entities in strategizing R&D initiatives, budget allocations, and outcome evaluations.Furthermore, this work provides a scaffold for assessing the contemporary status and prognosticating ZEB research's forthcoming trends.Nonetheless, certain limitations persist, such as potential miscategorization of R&D grants or human biases affecting thematic assessments.Pioneering language models, exemplified by OpenAI's ChatGPT or Google's Bard, could proffer resolutions in future research endeavors.
There exists an imperative for subsequent studies to explore the intrinsic attributes and the juxtaposition of R&D grant data with scientific publications.Such endeavors would accentuate the significance of R&D grant data, augmenting its analytical utility, and paving the way for diverse AI-driven scientometric evaluations.Our AI-enhanced assessment underscored the multifaceted nature of ZEB research, spanning domains from nuclear technology to material science.The emphasis on fields such as advanced nuclear technology heralds potential paradigm shifts in ZEB's energy and sustainability blueprints, signaling a renaissance in sustainable architectural design.
Based on the derived clusters, it is prudent for stakeholders to channel investments into emergent domains, such as cutting-edge materials, seismic safety paradigms, and novel nuclear innovations, as these could dictate ZEB's future trajectory.This work epitomizes AI's prowess in sifting through and categorizing intricate research vectors, suggesting that embracing such tools can refine the granularity and scope of scientific evaluations, thereby equipping stakeholders with actionable insights.Given the fluidity of ZEB research, a periodic reassessment of these trends becomes indispensable.A steadfast monitoring regimen, buttressed by state-of-the-art methodologies, is quintessential to ensure alignment with evolving technological advances and societal imperatives.A meticulous exploration of anomalies or outliers could also proffer a visionary perspective on the research frontier of ZEB.
By comprehensively mapping the ZEB terrain through AI, this study offers indispensable insights for a broad audience, ranging from researchers and policymakers to industry frontrunners.The ever-evolving tapestry of ZEB research necessitates sustained scrutiny and recalibration to propel sustainable and trailblazing architectural innovations.
• ((net OR nearly) AND (zero) AND (energy OR carbon OR emission) AND (build* OR hous* OR construction OR home*)) OR ((zero) AND (energy OR carbon OR emission) AND (build* OR hous* OR construction OR home*)) OR ((energy) AND (plus OR ultralow OR ultra-low) AND (build* OR hous* OR construction OR home*)).

Figure 1
Figure 1 displays the number of R&D grants associated with ZEB from 2000 to 2022.As illustrated in Figure1, grants related to ZEB have exhibited an overall increasing trend throughout the period.Notably, R&D grants related to ZEB demonstrated a substantial surge from 2019 to 2020.
• ((net OR nearly) AND (zero) AND (energy OR carbon OR emission) AND (build* OR hous* OR construction OR home*)) OR ((zero) AND (energy OR carbon OR emission) AND (build* OR hous* OR construction OR home*)) OR ((energy) AND (plus OR ultralow OR ultra-low) AND (build* OR hous* OR construction OR home*)).

Figure 1
Figure 1 displays the number of R&D grants associated with ZEB from 2000 to 2022.As illustrated in Figure1, grants related to ZEB have exhibited an overall increasing trend throughout the period.Notably, R&D grants related to ZEB demonstrated a substantial surge from 2019 to 2020.

Figure 1 .
Figure 1.Trend in the number of research and development (R&D) grants by start year, 2000-2022.

Figure 1 .
Figure 1.Trend in the number of research and development (R&D) grants by start year, 2000-2022.

For a term
x within class c W x, c = t f x, c × log 1 + A f x t f x, c = f requency o f word x in class c f x = f reqeuncy o f word x across all classes A = average number o f words per class (1)

Figure 2 .
Figure 2. Artificial intelligence-based process of collecting and analyzing related research and development grant data related to zero-energy building.

Figure 2 .
Figure 2. Artificial intelligence-based process of collecting and analyzing related research and development grant data related to zero-energy building.

Figure 3 .
Figure 3. Results of document embedding and clustering for research and development (R&D) grants related to zero-energy building.

Figure 3 .
Figure 3. Results of document embedding and clustering for research and development (R&D) grants related to zero-energy building.

Figure 4 .
Figure 4. Results of topic modeling by clusters.Figure 4. Results of topic modeling by clusters.

Figure 4 .
Figure 4. Results of topic modeling by clusters.Figure 4. Results of topic modeling by clusters.
0713583) Transportation (0.0703829) iFuelActive-Smart diesel fuel solutions for the low carbon transition (UK) Development of a new forest monorail using potential energy (Japan) Zero emission hauler (Sweden) Havyard-Zero-emission ROPAX vessel (Norway) Cluster_15 Fuel (0.1070516) Cell (0.0928153) Power (0.0597689) Energy (0.0475239) Sofc (0.0410947) Development of a retrofittable dry low-emissions industrial gas turbine combustion system for 100% hydrogen and natural gas blends (US) Collaboration to develop manufacturing methods of electric microreactors for clean hydrogen production (Canada) Safe, low-cost hydrogen storage materials from NZ resources (New Zealand) supercritical carbon dioxide oxy-combustor development and testing (US) SCC-CIVIC-PG Track A: Novel fuel-flexible combustion to enable ultra-clean and efficient waste-to-renewable energy in changing climate (US) Full-field laser vibrometry for combustion diagnostics (device (UK) Wave-energy converter performance and cost optimization through novel controls strategies (US) Development of micro water generator for household water pipes (Japan) Advanced modeling and simulation development of hydroelectric power generators, including electronic excitation circuits (Canada) Cluster_18 Material (0.0973246) Building (0.0820252) Fiber (0.0768128) Polymer (0.0720646) Insulation (0.0710307) Development, characterization, and study of the durability of flexible polymer eco-composites based on milkweed fibers for the building envelope (Canada) DL: Systems analysis and fundamental control of bacterial processes in the production of bio-concrete for construction purposes BioZEment 2.0 (Norway)

Table 1 .
Top 10funder countries for research and development (R&D) grants related to zero-energy building.

Table 1 .
Top 10 funder countries for research and development (R&D) grants related to zeroenergy building.Canada provided the most funding with 431 cases, followed by Innovate UK in the UK with 413 cases, the EC in Belgium with 312 cases, the Engineering and Physical Sciences Research Council in the UK with 225 cases, and the National Natural Science Foundation of China in China with 194 cases (Table 2).
1Belgium is the seat of the European Commission (EC) and includes R&D grants funded by the EC.In terms of funding institutions, the Natural Sciences and Engineering Research Council in

Table 2 .
Top 10 funder and funder countries for research and development (R&D) grants related to zero-energy building.

Table 3 .
Top 10 funder countries and average funding amount in US dollars for research and development (R&D) grants related to zero-energy building.

Country The Number of Funded R&D Grants Average Funding Amounts in USD
1Belgium is the seat of the European Commission (EC) and includes R&D grants funded by the EC.

Table 4 .
Sizes of 25 clusters from documents on research and development (R&D) grants related to zero-energy building.

Table 4 .
Sizes of 25 clusters from documents on research and development (R&D) grants related to zero-energy building.

Table 5 .
Topics in 25 clusters and titles of major research and development (R&D) grants related to zero-energy building (ZEB).

Table 6 .
Topics for 25 clusters and the titles of major research and development (R&D) grants related to zero-energy building (ZEB).Research fields include developing nuclear coolant systems, new fuel assemblies, and small modular reactor technology.Other research areas include advanced diagnostics for fusion energy R&D, high-fidelity digital twins for critical systems, and plasma focus generators for material research in nuclear fusion.Additionally, the grants cover innovative solutions for nuclear waste containment, cost reduction of advanced reactor operation and maintenance, and integral benchmark evaluations of zero-power tests and multicycle depletion experimental data.

Table 6 .
Cont. scale energy use with occupant behavior uncertainty, developing intelligent net-zero energy modular homes for cold regions, advancing the circular economy potential of waste-to-value processes, achieving near zero and positive energy settlements, and developing modeling and assessment capabilities to optimize the design and production of sustainable home and personal care consumer products to meet net-zero carbon targets.